March 2007 doc.: IEEE 802.22-07/0063r2 IEEE P802.22 Wireless RANs Status of PHY parameters discussion Date: 2007- 03-07 Author(s): Name Gerald Chouinard Company Communications Research Centre, Canada Address 3701 Carling Ave. Ottawa, Ontario Canada K2H 8S2 Phone email 613-998-2500 gerald.chouinard@crc.ca Abstract This contribution tries to summarize the status of the discussion on the 802.22 PHY parameters as of the end of the January 2007 session in London, U.K. for the purpose of focussing the discussion that will take place over teleconference calls between the two sessions so that finaldecisions can be made before or at the Orlando March session so that an agreed set of PHY parameters can be include in the Draft 1.0 for Working Group Balloting. The state of discussion on each key PHY parameter is given along with some observations from the perspective of the author and a “way forward” is proposed to try to accelerate the consensus process. Notice: This document has been prepared to assist IEEE 802.22. It is offered as a basis for discussion and is not binding on the contributing individual(s) or organization(s). The material in this document is subject to change in form and content after further study. The contributor(s) reserve(s) the right to add, amend or withdraw material contained herein. Release: The contributor grants a free, irrevocable license to the IEEE to incorporate material contained in this contribution, and any modifications thereof, in the creation of an IEEE Standards publication; to copyright in the IEEE’s name any IEEE Standards publication even though it may include portions of this contribution; and at the IEEE’s sole discretion to permit others to reproduce in whole or in part the resulting IEEE Standards publication. The contributor also acknowledges and accepts that this contribution may be made public by IEEE 802.22. Patent Policy and Procedures: The contributor is familiar with the IEEE 802 Patent Policy and Procedures <http://standards.ieee.org/guides/bylaws/sb-bylaws.pdf>, including the statement "IEEE standards may include the known use of patent(s), including patent applications, provided the IEEE receives assurance from the patent holder or applicant with respect to patents essential for compliance with both mandatory and optional portions of the standard." Early disclosure to the Working Group of patent information that might be relevant to the standard is essential to reduce the possibility for delays in the development process and increase the likelihood that the draft publication will be approved for publication. Please notify the Chair <Carl R. Stevenson> as early as possible, in written or electronic form, if patented technology (or technology under patent application) might be incorporated into a draft standard being developed within the IEEE 802.22 Working Group. If you have questions, contact the IEEE Patent Committee Administrator at <patcom@ieee.org>. Submission page 1 Gerald Chouinard, CRC March 2007 doc.: IEEE 802.22-07/0063r2 Status of the PHY parameters discussion and proposed way forward This document tries to summarize the key items that are left to be discussed before a final set of parameters for the PHY layer of the 802.22 WRAN standard can be adopted. For each item, a summary of the discussions so far and some observations from the author are given and a way forward is proposed. These key items are listed in order of priority for the decisions since the earlier items will need to be known to take appropriate decisions on the later ones. 1. Synchronization of the CPE The CPEs have to synchronize quickly with the base station and keep their time and frequency synchronization during active operation. When a CPE initially associates with a base station, some delay in synchronization can be acceptable (say less than one second). This allows a CPE to progressively converge on a proper time and frequency synchronization based on a number of frame headers (or super-frame headers). Once the synchronization in time and frequency has been achieved, the CPE needs to maintain this synchronization within the required tolerances. If the synchronization is lost, re-synchronization needs to be achieved within a delay acceptable for the types of services to be provided. The most demanding service will likely be real-time applications such as VoIP which requires a maximum latency of 20 ms. This means that the re-synchronization will need to be achieved within two frames of 10 ms. Observations: ETRI’s simulations demonstrated that synchronization can be acquired with one preamble symbol containing 2 PN-sequences. Monisha Gosh noted that this may need a rather complex synchronization scheme and synchronization robustness could be improved by using a super-frame preamble symbol containing 4 PN-sequences to increase the capture range of the synchronization algorithm. ETRI’s simulations include synchronization based on both fractional as well as integer subcarrier spacings, therefore claimed not to be limited in capture range. The use of a more robust preamble at the super-frame header would not help to maintain the synchronization for real-time applications like VoIP since its repetition would only occur every 16 frames. Way forward: A decision has to be taken as to whether a single symbol frame preamble is sufficient or two preamble symbols will be needed. As noted above, more robust synchronization at the super-frame level will not help. The trade-off is between system capacity overhead, synchronization robustness and synchronization scheme complexity. Efforts should be made to keep only one option to keep the first draft standard as simple as possible. Note that there is no need for such synchronization requirement on the upstream since it is assumed that once the CPE is in lock with the base station, its burst received at the base station will also be synchronous with the base station operation. Consensus was reached on Feb. 1st that a single preamble symbol per frame will be sufficient. The discussion on the need for a more robust preamble (one 4-sequence symbol and two 2-sequence symbols) for the super-frame is still open. More evidence is needed on its need given that it would not help real-time streaming such as VoIP but may allow more robust synchronization at edge of coverage and represents 0.7% loss in system throughput. See also comment #491, p. 58. 2. Frame preamble structure In order to achieve proper time and frequency synchronization, different PN-sequences are proposed for the frame preamble symbol. A repetition of two PN-sequences allows every other carriers to be modulated with null carriers in between to acquire a tight frequency synchronization. If a repetition of Submission page 2 Gerald Chouinard, CRC March 2007 doc.: IEEE 802.22-07/0063r2 three PN-sequences is used as in the case of the 802.16e standard, one over three carriers is modulated, allowing a somewhat less tight synchronization. A repetition of 4 PN-sequences will modulate one carrier over 4, resulting in somewhat looser frequency synchronization but it is claimed that the capture range would be better and would allow for simpler synchronization schemes. Another aspect is the fact that if more carriers are modulated in the preamble, more energy will be available at the CPE to synchronize. Observations: The ETRI simulations were done assuming a single preamble symbol with 2 PNsequences. Eli Sofer claims that a repetition of 3 PN-sequences as used in 802.16e shoud be used whereas Monisha Gosh claims that a preamble symbol with 4 PN-sequences followed by a preamble symbol with 2 PN-sequences would be needed. Way forward: After a decision has been taken on the number of frame preamble symbols needed as indicated in section 1, a decision on the number of PN-sequences has to be taken on evidence of the trade-offs involved. (Comment: #96, p.58) Once this is done, then, a decision has to be taken on the actual sequences to be used (actual sequences or initial vectors and generation polygons, e.g., 802.16e sequences vs. binary m-sequences as proposed by Monisha Gosh), the number of orthogonal sequences required, the acceptable level of PAPR and cross-correlation. (Comment: #58, p.57) Edward Au presented doc. #070002r2 on polyphase CAZAC sequences for low PAPR frame preambles. Larger is the decimation (2,3,4), higher is the PAPR. Edward Au, Wai Ho, Monisha Gosh and Eli Sofer to develop a table comparing the PN-sequences from 802.16, M-binary and polyphase to compare complexity in generating and processing for F & T sync and channel estimation, PAPR and crosscorrelation. [8 Feb.07] Eli: 3 PN sequences: three repetition of the PN sequence is proposed since it will occupy 1 over 3 carriers in the preamble symbol, thus allowing the transmission of three simultaneous preamble sequences for the three sectors using the three different subcarrier sets. Three different sets of FHC and US-DS-MAP will be needed for the three sectors. Eli proposed to concatenate the information for the three sectors in TDM resulting in the need for 2 FCH and Mapping symbols. [15 Feb.07] Three step decision process: - Decide on the number of PN sequence repetitions in the preamble symbol (2 or 3) - Decide on the type of PN sequences (CAZAC or binary) - Decide on which specific set of sequences and the number of orthogonal sequences needed. [22 Feb.07] Eli continued his presentation of the rationalization for the three sectors implementation, showing that 4 sectors is not good. He also reviewed the ranging process and presented the results of simulations on the 3x3 tiling structure which are not very good. Averaging over subsequent frames or selecting the better areas of the channels for allocating the subcarriers (AMC) would be needed to improve the performance. The 7x1 tiling seems to be more appropriate for WRAN channels. [28 Feb.07] Wai Ho Mow presented the latest version of the CAZAC polyphase sequence for reduction of the preamble PAPR. The question is whether the preamble can be boosted or not. An email was sent to Carl to clarify the averaging period over which the 4 W EIRP limit will be measured. The depth of interleaving for DTV is about 150 usce. The peaks should therefore be less than 150 usec. Thus the averaging would likely be at the symbol period. If that is the case, there is no room for boosting the preamble and the extra PAPR reduction provided by CAZAC sequences would not be needed. The 1/3 sectorized preamble may have an impact on the synchronization reliability. Also, 1 over 3 sub-carriers Submission page 3 Gerald Chouinard, CRC March 2007 doc.: IEEE 802.22-07/0063r2 in the frequency domain needs more than a simple 3 repetition PN-sequence. Ramon to ask Motorola on their experience with the 3 PN-sequence in implementing 802.16e. [8 March 07] Eli and ETRI to present results on 3-sequence. Debrief on 2 versus 3 repetition sequence and vote taken. Debrief from both sides on CAZAC versus Binary sequences and vote to be taken. 3. Channel training In order for the WRAN system to track the transmission channel variations, channel training information needs to be included as part of the symbols to be transmitted. Two trends exist in providing such training information. One is based on what is being done in 802.11 with the transmission of training symbols at the beginning of a burst whereas the other one is based on what is being done in 802.16 with the transmission of pilot carriers spread over a number of symbols. The amount of overhead information that is needed to allow reliable transmission will depend on the characteristics of the transmission channel, i.e., its time spread and frequency spread characteristics. The channel training will have different requirements in the downstream direction than in the upstream direction. The signal received by a CPE from a base station contains all the carriers of the multiplex whilst the signal received by the base station in the upstream direction only contains a subset of these carriers that can be spread over the entire 6 MHz channel. Proper interpolation between carriers can be done to fully characterize the channel in the downstream whereas this is not possible in the upstream direction. Note that even if only training symbols were to be used at the beginning of transmission bursts, some pilot carriers would still be needed within the burst to track the phase of the carriers due to phase noise in the receiver. Observations: The state of discussions on this matter is summarized in the spreadsheet: 22-06-026402-0000_OFDMA_Parameters.xls. It was assumed that the channel training in both downstream and upstream directions would be based on 240 pilot carriers per symbol, allowing all carriers to be visited over 7 consecutive symbols. It had also been assumed that a more solid channel training would be achieved at the super-frame level by a repetition of two training symbols in the superframe header, each containing 2 PN-sequences. As mentioned before, it seems that in order to meet the QoS level for realtime applications such as VoIP, such more robust training sequence at each super-frame will not be sufficient. An interesting situation arose in London where 802.11 tenants claimed that robust synchronization and channel training schemes based on preamble symbols would be needed while the 802.16 tenants were claiming that was not necessary and that less robust schemes such as the one used in 802.16 based on spread pioot carriers should be sufficient. Should 802.22 adopt a very robust synchronization and channel training schemes based on preambles used for “fixed” operation in 802.11 or less robust schemes that seem to be sufficient for “mobile” operation in 802.16? It would be useful here to have an assessment of the relative robustness on 802.11 versus 802.16 systems in a same 802.22 channel environment to see their relative performance. See comments #494, p.59, #144, p.60 and #495, p. 59. Way forward: A decision has to be taken as to the need for the robust 2-training symbol sequence à la 802.11 in the super-frame header. This 2-training symbol sequence had some support at the super-frame level because the overhead would have been small but if it is to appear at the beginning of each frame, the overhead would no longer be negligible. The ETRI simulation results showed that the 240 pilot carriers per symbol were sufficient and they indicated that in order to improve the transmission Submission page 4 Gerald Chouinard, CRC March 2007 doc.: IEEE 802.22-07/0063r2 granularity, only 3 symbols out of the 7 symbols that would be needed to completely recover all the pilot carriers could be acquired with a loss of 0.7 dB in performance. Such loss would not be as great in the downstream direction since the preamble symbol can be used to improve the channel estimation and that interpolation in the frequency domain can be done. The use of pilot carriers spread over 7 symbols rather than training symbols at the beginning of the bursts will result in a latency of 7 * 336= 2.35 ms accompanied with the corresponding buffering at the reception. A final decision will be needed between the training symbols alternative and the spread pilot carriers. There is no clear reason why the two options should be maintained in the initial draft standard if they both give equivalent channel training information. 4. Upstream symbol/carrier tiling In the case where the previous item is resolved in the direction of the use of pilot carriers spread over a number of symbols for channel estimation, the exact number of symbols and adjacent carriers over which the upstream transmission takes place and, hence over which the local channel response can be interpolated, that is the tiling pattern, needs to be decided upon. The CPE will be programmed to use a succession of these tiles along with the pilot carriers located within these tiles to transmit its sub-channel information and provide the base station with the channel state information needed for proper demodulation. Observations: The tile size that was assumed in the ETRI simulations and that is used in the OFDMA Parameters spreadsheet is 7 symbols by 1 carrier. This tile contains one pilot for a ratio of pilot to data capacity of 1/6. Eli Sofer has proposed the consideration of a 3 symbols by 3 carriers tile corresponding to the 802.16e upstream OPUSC option which gives a 1/8 pilot to data capacity. These two tile sizes need to be evaluated with the four channel models contained in document 22-05-0055-070000_WRAN_channel _Modeling.doc before conclusions can be reached on their relative merit. Simulation results are expected from ETRI and Runcom. See comment #13, p. 39, #149, p. 60. Way forward: Once the results of the simulations are known, the group should be in a position to take a decision based on the best fit of the tile size with the transmission channel time and frequency spreading characteristics expected in the TV bands. A proper trade-off between this best fit, the pilot carrier overhead and the transmission granularity should be found to arrive at the final decision. [8 March 07] John Benko will propose a 1X5 tile for upstream. 5. Subchannelization: carrier allocation pattern In the initial joint proposal, two main carrier allocation patterns were proposed: - distributed carrier permutation - adjacent carrier permutation (frequency selective diversity) The first one randomizes the carriers assigned to the different CPEs over the 6 MHz channel whereas the second one can take advantage of the channel state information to allocate the carriers in the portion of the 6 MHz channel that has the best performance for each CPE, therefore allowing a potential improvement of the overall transmission performance. Observations: There is however a drawback to the adjacent carrier permutation in the upstream direction because of the possible concentration of CPE power over a 200 kHz bandwidth which corresponds to the bandwidth that will affect to interference into Part 74 wireless microphones. The Submission page 5 Gerald Chouinard, CRC March 2007 doc.: IEEE 802.22-07/0063r2 difference between the carriers transmitted by a CPE being concentrated in 200 kHz rather than being spread over the entire TV channel corresponds to a worsening of the potential interference of some 10 dB at the edge of the WRAN coverage when the CPE tries to transmit its rated maximum capacity of 384 kbit/s using a 4 W EIRP. See comments 495, p.59, 145, 149, 151, 150, 147, p.60, 152, 146, p.61. Way forward: This 10 dB worsening of the potential interference towards the Part 74 wireless microphone incumbents needs to be considered carefully. This would tend to favor a simple randomization of the carriers transmitted from a CPE across the TV channel. The question to be addressed is whether the first version of the 802.22 standard should be limited to the distributed carrier permutation on both upstream and downstream for simplicity sake. It would be useful at the time of the decision to have indications of the extent of improvement that the adjacent carrier permutation could provide on the downstream and on the upstream as well as the extra MAC messages that would be required for proper signaling and operation such as the transmission of the channel state information from the CPE to the base station, and the complexity of the scheduling optimization needed at the base station to share the channel capacity among all active CPEs while taking advantage of the potentially better performance of all these channels at the same time. 6. Data to Carrier time/frequency interleaving schemes During the London session, Patrick Pirat raised the point about clarifying the detailed mapping of the data bits in terms of randomization over symbols and carriers. If such randomization is done in two steps, the randomization achieved through step one may be defeated by the second step of of the process. The standard therefore needs a fully understood and described data mapping in terms of time and frequency interleaving schemes. Draft 1.0 should contain the details of the sub-channel block sizes per frame and a detailed description of the interleaving schemes. Observations: It seems that this aspect has not been fully studied even though some assumptions must have been made to be able to simulate the performance of the various sets of PHY parameters in actual multipath channel conditions. It was noted that proper bit interleaving to map to the frequency time multiplex is needed for cyclic FEC codes while LDPC would already include such interleaving. The effect of truncating sub-channel capacity in order to increase the system granularity as suggested by ETRI (3 symbols instead of 7) needs to be studied as to its impact on the data randomization. See comment #144, #145, #147, #150 and #151, Page 60. Way forward: Work on this item should take place between now and the March session in Orlando so that a proper description of the interleaving schemes (data bit mapping to frequency/time physical multiplex) can be inserted in the Draft 1.0 in time for the WG balloting. Someone (Patrick Pirat?) should take the lead on this and form a small specialized ad-hoc group to discuss this matter and come up with a detailed section to be added to the Draft 1.0 in March. 7. Forward Error Correction The current Drat 0.2 includes the convolutional coding and Viterbi decoding scheme with puncturing for the inner FEC. [8 March 07] . No outer code is needed.. A motion was passed in London to include the consideration of advanced inner coding FEC schemes in the preparation of Draft 1.0. The joint proponent team had identified three other FEC schemes: - LDPC - Shortened block turbo code Submission page 6 Gerald Chouinard, CRC March 2007 - doc.: IEEE 802.22-07/0063r2 Duo-binary convolutional turbo code (CRSC) Observations: An ad-hoc group was formed in January 2006 during the Hawaii session to evaluate the relative merits of each of these FEC schemes but the pace has been slow due to the expectation that only the first scheme would be considered in the preparation of the first draft standard. Now that this has been put as a feature to be resolved in the first draft, simulation results are required urgently. Way forward: The work in the ad-hoc group needs to be accelerated a.s.a.p. and simulation results using the latest set of PHY parameters and the 4 channel models need to be produced before the March session so that a decision can be made in time for the Draft 1.0. To keep the standard simple, all efforts should be made to arrive at one FEC mode rather than multiple options. From comment resolution process (22-07-0063-00-0000_WGR2_Cmt_DB.pdf, page 45), once the selection of the FEC code is made, Table 29 of the draft standard will need to be updated. I2R does not agree with Carl's interpretation of the London motion which would only allow consideration of LDPC and turbo code and not shortened sequence as potions. If we go with the motion, the mandatory convolutional coding and the three optional codes would need to be included in the standard. [8 Feb.07] [8 Feb.07] [8 March 07] Possible presentation by John Benko if time permit. Bit interleaving may be affected by the type of FEC scheme. 8. Super-frame Need to address comments on the need of super-frames (#247, p.57). Need to identify the payload of the super-frame (one symbol at the beginning of the first frame in the series of 16 with a special PNsequence and one symbol less in this first frame to keep the frame length at 10 ms. Is a special superframe preamble needed at the receiver for frequency and time synchronization (initial burst detection, AGC tuning, coarse frequency offset estimation and timing synchronization)? What is the spreading method for the SCH? (Comments: #187, p.10, #93, p. 28, #525, #231, p.37, #179, #16, #17, p.41, #247, p 57, #493, #491, #96, p.58, #115, #250, p.61) 9. TDD versus TDD/FDD Should the first draft standard include what is necessary for a FDD operation? ETRI is to bring more supportive information on this topic. (Comment: 39, p.20, 252, p.62) [8 March 07] ETRI will distribute a document on FDD. 9. CBP PHY design Need for very robust synchronization and preamble generated at the transmitting CPE aimed at synchronizing another CPE that will receive this CBP upstream burst. Need to signal to some CPEs the need to monitor the channel with their WRAN receive chain during the upstream sub-frame and report on the received information. Need to specify a specific set of carriers to be used by the CBP to carry the CPE-to-CPE geolocation information. Comment: 191, p.14, 504, p.21, 217, 213, 215, p.23, 219, 221, 207, p.24, 220, 208, 210, 211, p.25, 212, 42, 100, p.26, 216, p.27, 46, p.33, 53, p.35, 87, 11, p.39, 14, 15, p.40, 180, p.43, 184, p.49) 10. Ranging methods Submission page 7 Gerald Chouinard, CRC March 2007 doc.: IEEE 802.22-07/0063r2 See comment 336, p.59. 11. Power control method See comments #137, p.8, #343, #509, p.9, #337, p.59. 12. OFDMA subchannellization See comment: 12, 13, p.39. 13. WRAN system performance at edge of coverage The system is designed to provide the required capacity as stated in the RFD, that is 384 kbit/s at the edge of coverage with 4 W maximum EIRP. In the WRAN reference model spreadsheet (22-04-000215-0000_WRAN_Reference_Model.xls), this is obtained by setting the required return bit rate at 384 kbit/s and finding the required minimum field strength at the base station receiver for a given set of receiver assumptions and a Eb/No of 5.6 dB which corresponds to QPSK, rate= 1/2 performance in a Gaussian channel. The distance is then found using the ITU-R P.1546 propagation model for 4 W transmit power from the CPE for a service availability of F(50, 99.9). This results in a reach of 31 km for a 75 m base station antenna height and a 10 m CPE antenna height. For the set of current OFDMA parameters (22-06-0264-02-0000_OFDMA_Parameters.xls), if the entire frame payload is assigned to the upstream capacity (i.e., 27 symbols for CP= 1/8), 389 kbit/s can be transmitted using 6 sub-channels, that is 168 carriers, 1/10th of the total number of carriers in the OFDMA multiplex. Lower capacity can still be achieved from further away CPEs until it hits the capacity of a single sub-channel (i.e., since transmission cannot be reduced to a fraction of a subchannel), that is an additional propagation marging of 7.8 dB for a capacity reduced to 64.8 kbit/s. With respect to the downstream, if the base station EIRP is limited to 4 W, the shared downstream payload at the limit of coverage will be 389 kbit/s transmitted on 168 QPSK modulated carriers occupying the total 27 symbols in a frame if the data has to be sent to CPEs located at the edge of the coverage area because of the reciprocal RF path, and this shared payload could decrease with distance beyond the edge of coverage similar to what is described above for the CPE. If the base station EIRP is allowed to increase beyond 4 W, then the entire set of OFDM carriers (i.e., 10 times more) could be transmitted to serve more CPEs at the edge of coverage. A 40 W base station EIRP would allow the transmission of 3.888 Mbit/s toward CPEs at the edge of the coverage area if the frame is fully allocated to downstream traffic. In response to comment: #483, p.56, #494, p. 59.14. Other comments from London session [1] Proposals accepted / Need follow-up Comment #484, Page 57; Comment #492, Page 58 Comment #152, Page 61 Comment #148, Page 61 [2] Accept in Principle / Defer under discussion Comment #483, Page 56 - discussion on more robust coding techniques and include remark about EIRP and BS height. Comment #336, Page 59 - discussion on ranging. ___________________________ Submission page 8 Gerald Chouinard, CRC